Warm-up Periods

Indicators like EMA, RSI, and ATR need a minimum number of bars before they produce valid output. The warm-up period prevents your strategy from trading on incomplete data.

The Problem

On the first few bars, indicators return NaN because they haven't accumulated enough data:

Bar 0:  EMA(20) = NaN
Bar 1:  EMA(20) = NaN
...
Bar 19: EMA(20) = NaN
Bar 20: EMA(20) = 48523.45  (first valid value)

If your strategy doesn't handle this, it will make decisions based on invalid data.

How warmupBars Works

The warmupBars parameter tells the engine to skip the first N bars of onBar() execution entirely. Your strategy code simply won't run until bar N.

// In your strategy parameters
{
  warmupBars: 50  // Skip the first 50 bars
}

The engine starts the onBar loop at bar warmupBars instead of bar 0. This is more efficient than checking q.isNaN() on every bar — the engine skips the loop iterations entirely.

Calculating the Right Value

Your warm-up should be at least as long as the longest indicator period, plus a safety margin:

Rule of thumb: Take the largest indicator period and add 1.

// If your strategy uses EMA(50) and RSI(14):
// Warm-up = max(50, 14) + 1 = 51
{
  warmupBars: 51
}

Defensive Coding: q.isNaN()

Even with warmupBars set, it's good practice to guard against NaN values. This protects against edge cases and makes your strategy robust:

function onBar(ctx, i) {
  const ema = ctx.ind.ema[i];
  const rsi = ctx.ind.rsi[i];

  // Early return if any indicator isn't ready
  if (q.isNaN(ema) || q.isNaN(rsi)) return;

  // Safe to use indicators here
  if (rsi < 30 && q.crossOver(ctx.series.close, ctx.ind.ema, i)) {
    ctx.order.market('ASSET', 1, { signal: 'buy' });
  }
}

Chained Indicators

When one indicator depends on another's output, warm-up compounds:

function init(ctx) {
  ctx.addIndicator('ema', 'EMA', { period: 50 });
  ctx.addIndicator('rsi', 'RSI', { period: 14 });
  // If you use EMA of RSI, you'd need 50 + 14 = 64 bars
}

Impact on Backtests

• Warm-up bars consume data but don't generate trades
• A 200-bar warm-up on 500 bars of data leaves only 300 bars for actual trading
• Shorter timeframes (1h) provide more bars, so warm-up is less impactful
• Longer timeframes (1d) with large warm-ups may leave too few bars — use adequate date ranges

Common Mistakes

1. Setting warmupBars too low

// Wrong: EMA(200) needs at least 200 bars
{ warmupBars: 50 }

// Correct
{ warmupBars: 201 }

2. Not accounting for all indicators

// Uses EMA(20), RSI(14), and BBANDS(50)
// warmupBars should be max(20, 14, 50) + 1 = 51

3. Forgetting compound indicators MACD uses a 26-period EMA and a 9-period signal line, so it needs at least 35 bars.

Related

• init() — Declaring indicators
• q.isNaN() — Checking indicator readiness
• Indicator Computation — How the engine computes indicators